def MakePlot(mask,title,filename):
fig=plt.figure(figsize=(16, 12))
fig.subplots_adjust(wspace=.15,right=.95, left=.05,top=.93, bottom=.05, hspace=.35)
for iface_index,iface_name in enumerate(['X','Y','Z','XZ','XY','YZ']):
ax = fig.add_subplot(3,2,iface_index+1)
import commands
files=commands.getoutput(mask %iface_name).split()
for index, file in enumerate(files):
residues=file.split('_')
residues=[i for i in residues if 'ASU' in i]
residues=residues[0].split('ASU')
data=genfromtxt(file)
plot_label=[i for i in file.split('_') if 'ASU' in i][0]
ax.plot(data[1:,0]*0.2,smooth(data[:,1],50),colors[index],linewidth=2,label=plot_label)
ax.plot ([1,data[-1,0]*0.2],[data[0,1],data[0,1]],'k',linewidth=2)
for label in ax.xaxis.get_ticklabels():
label.set_fontsize(8)
for label in ax.yaxis.get_ticklabels():
label.set_fontsize(8)
plt.title(title %iface_name,fontsize=12)
plt.xlabel('Time (ns)',fontsize=8, labelpad=0)
#ax.set_xticklabels([0,1.0,2.0,3.0,4.0,5.0])
plt.ylabel(r"Distace ($\AA$)",fontsize=8, labelpad=0)
#~ plt.ylim((0,6))
#plt.xlim(xmax=51) #modify to trajectory length
ax.yaxis.set_ticks_position('left')
ax.xaxis.set_ticks_position('bottom')
for line in ax.get_xticklines() + ax.get_yticklines():
line.set_markeredgewidth(4)
line.set_markersize(10)
from matplotlib.font_manager import fontManager, FontProperties
font=FontProperties(size=8)
if iface_index==5:
ax.legend(bbox_to_anchor=(0, 0, 1.1, .3),prop=font,ncol=3)
#~ plt.show()
plt.savefig(filename)
totalp=0.0
contactp=0
print "crystal contacts:"
for i in range(len(contact_names)):
name=contact_names[i]
data=all_data[:,i]
totalp+=average(data)
if data[-1]==ASUS:
print name
is_cryst=1
else:
is_cryst=0
if average(data) <4 and is_cryst!=1:
continue
fig=plt.figure(figsize=(16, 12))
plt.ylim((0,ASUS+1))
ax = fig.add_subplot(111)
ax.plot(smooth(data,10),'k')
plt.title('%s' %name, fontsize=24)
fig.suptitle('%s' %iface, fontsize=18)
plt.legend(["%5.2f" %average(data)])
contactp+=1
if is_cryst:
plt.annotate('crystal contact',xy=(5000,12.1), xytext=(5000,12.1),fontsize=18)
plt.savefig('residContact_plots_tmp/%s/%s.png' %(iface,name))
#~ totalp=totalp/contactp
#~ totalp=totalp/
print "%s %5.4f\n" %(iface, totalp)
p = 0.15 # edge probability
env = MVC(n,p)
cuda_flag = True
alg = DiscreteActorCritic(env,cuda_flag)
num_episodes = 4000
for i in range(num_episodes):
T1 = time.time()
log = alg.train()
T2 = time.time()
print('Epoch: {}. R: {}. TD error: {}. H: {}. T: {}'.format(i,np.round(log.get_current('tot_return'),2),np.round(log.get_current('TD_error'),3),np.round(log.get_current('entropy'),3),np.round(T2-T1,3)))
Y = np.asarray(log.get_log('tot_return'))
Y2 = smooth(Y)
x = np.linspace(0, len(Y), len(Y))
fig2 = plt.figure()
ax2 = plt.axes()
ax2.plot(x, Y , Y2)
plt.xlabel('episodes')
plt.ylabel('episode return')
Y = np.asarray(log.get_log('TD_error'))
Y2 = smooth(Y)
x = np.linspace(0, len(Y), len(Y))
fig2 = plt.figure()
ax2 = plt.axes()
ax2.plot(x, Y , Y2)
plt.xlabel('episodes')
plt.ylabel('mean TD error')
T1 = time.time()
log = alg.run_epoch()
T2 = time.time()
distances.append(log.get_current('final_dist'))
Y = np.asarray(distances)
Y[Y > 1] = 1.0
Y = 1 - Y
print('done: {} of {}. loss: {}. success rate: {}. time: {}'.format(
i, epochs, np.round(log.get_current('avg_loss'), 2),
np.round(np.mean(Y), 2), np.round(T2 - T1, 3)))
if (i % 100) == 0:
torch.save(alg.model.state_dict(), 'model.pt')
torch.save(alg.image_mean, 'norm.pt')
Y = np.asarray(log.get_log('final_dist'))
Y2 = smooth(Y)
x = np.linspace(0, len(Y), len(Y))
fig1 = plt.figure()
ax1 = plt.axes()
ax1.plot(x, Y, Y2)
plt.xlabel('episodes')
plt.ylabel('minimum episode distance')
Y = np.asarray(log.get_log('avg_loss'))
Y2 = smooth(Y)
x = np.linspace(0, len(Y), len(Y))
fig2 = plt.figure()
ax2 = plt.axes()
ax2.plot(x, Y, Y2)
plt.xlabel('episodes')
plt.ylabel('average loss')
#get experimental name and value
c_name = angle_names[angle]
c_exp = float(angle_exp[angle])
fig = plt.figure(figsize=(16, 12))
ax = fig.add_subplot(111)
ax.plot([0, 800], [0, 0], 'k')
#for each asu
for asu in range(12):
#get simulation value
c_sim = angle_sim[asu, ::10, angle]
#zero around experimental value
c_sim = c_sim - c_exp
#wrap values <-180 or >180
index = c_sim > 180
c_sim[index] = c_sim[index] - 360
index = c_sim < -180
c_sim[index] = c_sim[index] + 360
#apply hanning smoother
ax.plot(smooth(c_sim), colors[asu], label=asu + 1)
#~ ax.plot(c_sim,label=asu)
plt.title(c_name, fontsize=28)
ax.legend(bbox_to_anchor=(0, 0, 1.08, 1))
plt.xlabel('time', fontsize=28, labelpad=10)
plt.ylabel(r"degrees", fontsize=28, labelpad=10)
plt.ylim((-180, 180))
##~ plt.show()
plt.savefig('PerResidue/%s_%s.png' %
(c_name.split(':')[1], c_name.split(':')[0]))
print angle
print "i am done"
